yyashpatel / resnet_pytorch

Deep Residual Networks for Image Classification

Pytorch implementation of ResNet18 for CIFAR10 classification.

Authors

It's a joint work between Aditya, Vijay and Yash

• Aditya Wagh
• Vijayraj Gohil
• Yash Patel

Setup

Install either of Anaconda, Miniconda or Miniforge. All of them provide the conda package manager for python which we will use to install our project packages.

The code has beed tested using the following python packages.

• torch >= v1.10
• torchvision >= v0.11

Create a virtual environment to install project dependencies.

conda create -n torch

Activate the virtual environment.

conda activate torch

Install PyTorch and Torchvision using from the pytorch channel.

conda install pytorch torchvision matplotlib cudatoolkit=11.3 -c pytorch

Usage

Argument	Type	Default value	Description
-en or --experiment-number	required	1	Required. Your API key
-e or --epochs	required	120	Required. Your API key
-o or --optimizer	required	N/A	Required. Your API key
-d or --device	required	gpu	Required. Your API key
-lr or --learning-rate	required	0.1	Required. Your API key
-mo or --momentum	required	0.9	Required. Your API key
-wd or --weight-decay	required	5e-4	Required. Your API key
-dp or --data-path	required	./data	Required. Your API key
-b or --blocks	required	required	Required. Your API key
-c or --channels	required	required	Required. Your API key
'-m' or '--model'	required	'resnet'	Required. Your API key

python main.py \
    --experiment-number 1 \ 
    --optimizer sgd \ 
    --data-path ./data \
    --blocks 2 2 2 2 \
    --channel 54 96 188 324

See what these commands do.

usage: main.py [-h] -en EXPERIMENT_NUMBER -o OPTIMISER [-d DEVICE] [-e EPOCHS] [-lr LEARNING_RATE] [-m MOMENTUM] [-wd WEIGHT_DECAY] -dp DATA_PATH -b BLOCKS BLOCKS BLOCKS BLOCKS -c CHANNELS CHANNELS CHANNELS CHANNELS

  optional arguments:
  -h, --help            show this help message and exit
  -en EXPERIMENT_NUMBER, --experiment_number EXPERIMENT_NUMBER
                          number to track the different experiments
  -o OPTIMISER, --optimiser OPTIMISER
                          optimizer for training
  -m MODEL, --model MODEL
                          model to train
  -d DEVICE, --device DEVICE
                          device to train on
  -e EPOCHS, --epochs EPOCHS
                          number of epochs to train for
  -lr LEARNING_RATE, --learning-rate LEARNING_RATE
                          learning rate for the optimizer
  -m MOMENTUM, --momentum MOMENTUM
                          momentum value for optimizer if applicable
  -wd WEIGHT_DECAY, --weight-decay WEIGHT_DECAY
                          weight decay value for the optimizer if applicable
  -dp DATA_PATH, --data-path DATA_PATH
                          path to the dataset
  -b BLOCKS BLOCKS BLOCKS BLOCKS, --blocks BLOCKS BLOCKS BLOCKS BLOCKS
                          number of blocks in each layer
  -c CHANNELS CHANNELS CHANNELS CHANNELS, --channels CHANNELS CHANNELS CHANNELS CHANNELS
                          number of channels in each layer